82559 prefetches the next command

[Donald Becker]

On Sat, 15 Apr 2000, Kallol Biswas wrote:

>            I have very little knowledge about 82559 card and
> network programming. I was just reading your driver(not the latest one 
> I believe). 82559 prefetches the next command , won't there be 
> any problem if it prefetches the next command before the driver
> sets it up?

The 'suspend' bit on the last valid command prevents this from happening.

The suspend bit semantics are actually more complex than you migth expect in
order to make this work: the chip notes if the suspend bit is set the first
time it reads the descriptor.  After the chip has finished executing the
command the chip might need to re-read the descriptor to see if the suspend
bit has just been cleared.

> Question is what happens when the CU executes a command
> with invalid parameter, does it die?

Horribly.  Usually it can interpret the corrupt data as a command.
Sometimes it just stops.

> How do you determine that
> what fault it has encountered? By looking at Port Dump/Command
> dump registers?

Ehhhh, not really.  I've dumped the chip state, but it never reveals
anything useful.

> Is there any data manual for 82559's internal 
> micro machine registers like we have for most of the SCSI IO 
> cards?

No, that's highly secrete information.  Presumably the format might change
with each revision, even if the PCI version number isn't incremented.

BTW, v1.10 of the driver is available from
   ftp://scyld.com/pub/network/
You will also need
 Makefile
 pci-scan.[ch]
 kern_compat.h

Or you can just get the latest SRPM from the directory...

Donald Becker				becker@scyld.com
Scyld Computing Corporation
410 Severn Ave. Suite 210
Annapolis MD 21403


-------------------------------------------------------------------
To unsubscribe send a message body containing "unsubscribe"
to linux-eepro100-request@beowulf.org